This cohort study evaluates the association of antidiabetic medication initiation timing with the risk of major adverse cardiovascular events and all-cause mortality among individuals newly meeting diagnostic criteria for type 2 diabetes.
Key Points
Question
Is there a difference in the risk of major adverse cardiovascular events (MACE) and mortality among individuals who initiated antidiabetic medication (ADM) within 3, 6, or 12 months after meeting the diagnostic threshold for type 2 diabetes?
Findings
In this cohort study involving 23 452 participants, earlier ADM initiation after meeting the diagnostic threshold for type 2 diabetes was not associated with risk of MACE but was associated with a lower risk of mortality over a 5-year follow-up.
Meaning
These findings suggest health care authorities should adopt evidence-based strategies that promote timely intervention to optimize outcomes in type 2 diabetes management.
Abstract
Importance
Among individuals who meet the diagnostic threshold for type 2 diabetes (T2D), timely initiation of antidiabetic medication (ADM) is essential for lowering long-term cardiovascular risk.
Objective
To estimate the association between ADM initiation timing—specifically within 3, 6, or 12 months—and the risk of major adverse cardiovascular events (MACE) and all-cause mortality among individuals newly meeting diagnostic criteria for T2D.
Design, Setting, and Participants
This cohort study used target trial emulation to analyze health screening data linked to health insurance claims in Korea (2013-2022) using a clone-censor-weight approach. Participants were adults with newly detected glycated hemoglobin (HbA1c) of 6.5% or greater or fasting plasma glucose of 126 mg/dL or greater. Data analysis was conducted from January to August 2025.
Exposures
Eligible participants were cloned into 4 treatment strategies: ADM initiation within 3, 6, or 12 months or no initiation within 12 months (strategy 1, 2, 3, and control, respectively).
Main outcomes and Measures
Five-year absolute risk difference (RD) and risk ratio (RR) of 3-point MACE (stroke, myocardial infarction, and all-cause mortality) and all-cause mortality were estimated using Kaplan-Meier survival probabilities along with 95% CIs from 1000-sample nonparametric bootstrapping.
Results
A total of 23 452 eligible participants (mean [SD] age, 48.2 [11.1] years; 5790 [24.7%] female; mean [SD] HbA1c, 6.9% [1.1%]) were cloned into 4 treatment strategies. Earlier ADM initiation compared with the control showed progressively lower point estimates for 3-point MACE (RR, 0.32; 95% CI, 0.15 to 1.11 for strategy 1; RR, 0.65; 95% CI, 0.41 to 1.29 for strategy 2; RR, 0.93; 95% CI, 0.70 to 1.41 for strategy 3), though it did not achieve statistical significance. Corresponding RDs were −0.97% (95% CI, −1.26% to 0.14%), −0.49% (−0.84% to 0.40%), and −0.10% (−0.44% to 0.57%), respectively. ADM initiation within 3 months yielded significant risk reduction for all-cause mortality compared with the control in both relative (RR, 0.31; 95% CI, 0.10-0.98) and absolute (RD, −0.40%; 95% CI, −0.57% to −0.01%) scales.
Conclusions and Relevance
In this cohort study, earlier ADM initiation following the diagnostic threshold for T2D showed a lower risk of mortality, suggesting a potential cardiovascular benefit of early glycemic control; however, given the low event counts and the observational nature of the study, further evaluation in larger studies is warranted before definitive conclusions can be drawn.
Introduction
Type 2 diabetes (T2D) is a progressive metabolic disorder characterized by chronic hyperglycemia, which contributes to an increased risk of cardiovascular disease and mortality.1 Given that cardiovascular complications are the leading cause of death in individuals with T2D,2,3 mitigating these long-term adverse outcomes is a key therapeutic goal in diabetes care.4 Ideally, regular health screenings facilitate timely diagnosis and initiation of antihyperglycemic treatment based on glycemic status. However, T2D often remains undiagnosed5 for years due to mild or absent symptoms in its early stages, leading to delays in diagnosis and treatment.6
Previous studies have shown that intensive glycemic control in early diabetes can reduce major adverse cardiovascular events (MACE) and mortality.7,8 Notably, a 24-year posttrial follow-up of the UK Prospective Diabetes Study (UKPDS)9 found that improved glycemic control at diagnosis significantly reduced the risk of myocardial infarction and all-cause mortality, particularly among those achieving near normoglycemia within the first year—benefits that persisted even after later glycemic deterioration. This long-term benefit, attributed to the so-called “legacy effect”10 or “metabolic memory,”11 resulted from protection against early hyperglycemia-induced formation of advanced glycation end products and persistent inflammatory responses triggered by mitochondrial superoxide production. Based on this prior evidence underscoring the importance of early glycemic control, recent clinical guidelines advocate prompt pharmacologic intervention upon diagnosis of T2D.4
However, evidence regarding the outcomes of specific timing of antidiabetic medication (ADM) initiation on the risk of cardiovascular events and mortality remains limited, partly due to methodological challenges in observational studies, including immortal time and selection biases.12 To address these limitations, we conducted a retrospective cohort study using a clone-censor-weight approach to estimate the association between initiating ADM within 3, 6, or 12 months and the risk of MACE and all-cause mortality among individuals with newly diagnosed T2D.13
Methods
Data Source
We conducted a retrospective cohort study by linking the Kangbuk Samsung Health Study (KSHS) cohort with Korea’s National Health Insurance claims database from 2013 to 2022. The KSHS cohort includes adults (aged ≥18 years) who participated in comprehensive annual or biennial health checkups at the Kangbuk Samsung Hospital Total Healthcare Center. The study population included individuals present in both datasets, enabling integration of unique and complementary information from both data sources. Details regarding this linked database are provided in eAppendix 1 in Supplement 1. This study was approved by the institutional review board of Sungkyunkwan University and followed the Transparent Reporting of Observational Studies Emulating a Target Trial (TARGET) guideline.14 The requirement for informed consent was waived since all the data were anonymized.
Study Population
We included individuals with prediabetes (glycated hemoglobin [HbA1c] level ≥5.7%) between January 1, 2013, and December 31, 2022. Only health screening results with no missing HbA1c or fasting plasma glucose (FPG) values were included in the analysis. Then, those with an HbA1c level of 6.5% or higher or an FPG level of 126 mg/dL or higher were included in the study cohort. The cohort entry date, or time 0, was defined as the date on which the first-ever HbA1c level of 6.5% or greater or FPG level of 126 mg/dL or greater was recorded. Each individual had at least a 1-year washout period to ensure this was their first time exceeding the threshold for T2D. Individuals with prescription of ADM or diagnosis of type 1 diabetes at any time before the cohort entry date were excluded.
Treatment Strategies
ADM including metformin, thiazolidinediones, sulfonylureas, sodium-glucose cotransporter 2 (SGLT-2) inhibitors, glucagon-like peptide 1 (GLP-1) receptor agonists, dipeptidyl-peptidase 4 (DPP-4) inhibitors, insulin, alpha-glucosidase inhibitors, and meglitinides were considered (eTable 1 in Supplement 1). Initiation of any of these medications was defined as treatment initiation. We defined 4 treatment strategies based on the timing of treatment initiation since time 0: initiation within 3 months (strategy 1), initiation within 6 months (strategy 2), initiation within 12 months (strategy 3), and control group for no initiation within 12 months (strategy 4).
Outcome Definition
The primary outcome was a composite of modified 3-point MACE including myocardial infarction, stroke, and all-cause mortality. Since the linked database did not provide the cause of death, we were not able to define cardiovascular death. We also assessed the all-cause mortality as a separate coprimary outcome. Definitions for myocardial infarction and stroke have been validated in the NHIS database with high positive predictive values (approximately 81%-92%; eTable 5 in Supplement 1). We followed each individual until the first of occurrence of an event, 5 years after baseline, or study end date (December 31, 2022).
Clone-Censor-Weight Approach
We prespecified a hypothetical pragmatic trial (eTable 6 in Supplement 1) and emulated it using a cloning, censoring, and weighting approach. This approach enabled us to compare different treatment strategies that are indiscernible at time 0.12 In the cloning step, eligible participants were replicated into 4 clones and were assigned to 1 of 4 treatment strategy groups. Then, to ensure that clones adhered to their assigned treatment strategy during the follow-up, individuals were censored if they deviated from the assigned treatment strategy (censoring step). Specifically, artificial censoring was applied at the end of the grace period for strategies 1 through 3: 3 months for strategy 1, 6 months for strategy 2, and 12 months for strategy 3. For strategy 4 (no initiation within 12 months), clones were censored at the time point when a patient initiated ADM. Since the deviation from assigned treatment might be driven by baseline and postbaseline prognostic factors, this artificial censoring could induce selection bias. Therefore, we estimated time-varying inverse probability of censoring weights at monthly intervals and weighted the data to adjust for this potential selection bias (weighting step; eAppendix 2 in Supplement 1). This approach targets the observational analogue of the per-protocol effect.
Covariates
We assessed age, sex, and cohort entry year on cohort entry date, as well as comorbidities, comedications, and proxies for health-seeking behaviors (number of hospitalizations and physician visits) within a year before cohort entry date. Clinical variables, education status, physical activity, smoking, and drinking behavior were assessed within 3 years before cohort entry date. All these variables were included as baseline covariates for the pooled logistic regression models calculating inverse probability censoring weights. Time-varying covariates, including age, cohort entry year, comorbidities, and comedications, were assessed monthly and lagged by 1 month. For the covariates with missing values, we designated an unknown category and then included them in the logistic model.15 Detailed information on all covariates is provided in eAppendix 3 in Supplement 1, and their definitions are listed in eTable 7 in Supplement 1.
Statistical Analysis
Baseline characteristics of the study cohort were presented using descriptive statistics. We also assessed the distribution of periods between cohort entry and treatment initiation. For each treatment strategy, we identified and reported the number of individuals who deviated from or adhered to the assigned treatment strategy. To adjust for selection bias induced by artificial censoring, we calculated inverse probability of censoring weights, with weights updated at the beginning of each calendar month throughout follow-up. The denominator of the weights was the probability that each individual remained uncensored during follow-up, conditional on follow-up time, baseline covariates and time-varying covariates. Age was modeled as a continuous variable using cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. To address extreme weights, the weights were stabilized with numerator as the probability of being uncensored in a separate pooled logistic regression model conditional on follow-up time and baseline covariates (eAppendix 2 in Supplement 1). After cloning, censoring, and weighting, we applied the weighted Kaplan-Meier estimator16 to estimate survival probabilities for MACE and all-cause mortality under each treatment strategy and calculated 5-year absolute risk (%) as 1 minus the survival probability. We also estimated 5-year risk differences (RDs) and risk ratios (RRs) between treatment strategies, using group 4 (no initiation within 12 months) as the reference. For all effect estimates, pointwise 95% CIs were calculated by using nonparametric bootstrapping based on 1000 full samples. The result was considered statistically significant if the CIs for the RRs did not include 1.
We also conducted subgroup analyses by stratifying eligible participants based on age group (<65 vs ≥65 years), sex, eGFR at baseline (<60 vs ≥60 mL/min/1.73m2), presence of cardiovascular diseases (yes vs no), cohort entry year (2013-2017 vs 2018-2022), and baseline glycemic status (higher vs lower). Higher baseline glycemic status was defined as HbA1c of 7.0% or greater or fasting plasma glucose 130 mg/dL or greater, with all remaining participants classified as the lower glycemic status group. Several prespecified sensitivity analyses were conducted to verify the robustness of main findings. These included excluding insulin from ADM while adjusting for time-varying insulin use, additional adjustment for time-varying proxies of health care-seeking behavior, truncation of weights at the 99th percentile, and additional censoring at ADM discontinuation. We also evaluated negative control outcomes and performed a landmark analysis as an alternative approach (eAppendix 2 in Supplement 1). Details on sensitivity analyses are provided in eAppendix 4 in Supplement 1. All statistical analyses were conducted with SAS software, version 9.4 (SAS Institute). Data analysis was conducted from January to August 2025.
Results
Baseline Characteristics of Eligible Participants and Distribution of ADM Initiation
As presented in Figure 1, 23 452 eligible participants who exceeded HbA1c or FPG thresholds were replicated into 4 clones and assigned to 1 of the 4 treatment strategies. The proportions of participants who adhered to the assigned treatment strategies were 3651 (15.6%) for strategy 1, 5057 (21.6%) for strategy 2, 6988 (29.8%) for strategy 3, and 18 521 (79.0%) for strategy 4. Descriptive results for baseline characteristics and for time to initiation and duration of ADM use among patients adherent to each strategy are shown in eTable 8 and eTable 9 in Supplement 1, respectively. At baseline, participants had a mean (SD) age of 48.2 (11.1) years, 5790 (24.7%) were female, 5629 (24.0%) had hypertension, and 4824 (20.6%) used ARBs. The mean (SD) HbA1c was 6.9% (1.1) and mean (SD) FPG was 141.8 (33.4) mg/dL on the cohort entry date (Table 1). Over the 12 months of grace period, 4932 of 23 452 participants (21.0%) had initiated ADM, whereas 12 496 (53.3%) remained untreated throughout the 5-year follow-up period (eTable 2, eFigure in Supplement 1). Among 4932 patients who initiated ADM within 12 months, DPP-4 inhibitors accounted for the largest proportion (5181 participants [33.1%]), followed by metformin (4816 participants [30.7%]), and SGLT-2 inhibitors (1849 participants [11.8%]) (eTable 3 in Supplement 1). Of the 1041 patients who initiated monotherapy rather than combination therapy, the majority initiated with metformin (936 participants [89.9%]), followed by DPP-4 inhibitors (44 participants [4.2%]) and insulin (28 participants [2.7%]) (eTable 4 in Supplement 1).
Figure 1. Flowchart for Study Participants Selection.
FPG indicates fasting plasma glucose; HbA1c, glycated hemoglobin.
Table 1. Baseline Characteristics of Study Participants.
| Baseline characteristicsa | Participants, No. (%) (N = 23 452) |
|---|---|
| HbA1c on the cohort entry date, mean (SD), % | 6.9 (1.1) |
| FPG on the cohort entry date, mean (SD), mg/dL | 141.8 (33.4) |
| Demographic information | |
| Age, mean (SD), y | 48.2 (11.1) |
| Age category, y | |
| 18-29 | 584 (2.5) |
| 30-39 | 4951 (21.1) |
| 40-49 | 7978 (34.0) |
| 50-64 | 7842 (33.4) |
| ≥65 | 2097 (8.9) |
| Sex | |
| Male | 17 662 (75.3) |
| Female | 5790 (24.7) |
| Education statusb | |
| ≥University | 5009 (21.4) |
| ≤High school | 3711 (15.8) |
| Unknown | 14 732 (62.8) |
| Physical activity, d/wkb | |
| 0 | 5296 (22.6) |
| 1-2 | 2362 (10.1) |
| 3-7 | 1187 (5.1) |
| Unknown | 14 607 (62.3) |
| Cohort entry year | |
| 2013 | 3226 (13.8) |
| 2014 | 2266 (9.7) |
| 2015 | 2119 (9.0) |
| 2016 | 1950 (8.3) |
| 2017 | 1985 (8.5) |
| 2018 | 2344 (10.0) |
| 2019 | 2433 (10.4) |
| 2020 | 2229 (9.5) |
| 2021 | 2506 (10.7) |
| 2022 | 2394 (10.2) |
| Smokingb | |
| Never | 9306 (39.7) |
| Past | 755 (3.2) |
| Current | 6338 (27.0) |
| Unknown | 7053 (30.1) |
| Drinkingb | |
| No | 9050 (38.6) |
| Yes | 7447 (31.8) |
| Unknown | 6955 (29.7) |
| Comorbidities | |
| Ischemic heart disease | 567 (2.4) |
| Atherosclerosis | 184 (0.8) |
| Stroke | 157 (0.7) |
| Transient ischemic attack | 72 (0.3) |
| Other cerebrovascular disease | 159 (0.7) |
| Heart failure | 156 (0.7) |
| Atrial fibrillation | 211 (0.9) |
| Hypertension | 5629 (24.0) |
| Hyperlipidemia | 4302 (18.3) |
| Liver disease | 1802 (7.7) |
| Osteoarthritis | 2050 (8.7) |
| Inflammatory arthritis | 3311 (14.1) |
| Osteoporosis | 366 (1.6) |
| Thyroid disease | 798 (3.4) |
| COPD | 401 (1.7) |
| Asthma | 744 (3.2) |
| Pneumonia | 868 (3.7) |
| Diabetic retinopathy | 723 (3.1) |
| Diabetic neuropathy | 416 (1.8) |
| Diabetic nephropathy | 32 (0.1) |
| Hypoglycemia | 6 (<0.1) |
| Medications | |
| ACE inhibitors | 89 (0.4) |
| ARBs | 4824 (20.6) |
| β Blocker | 1447 (6.2) |
| Calcium channel blocker | 4090 (17.4) |
| Loop diuretics | 184 (0.8) |
| Thiazide | 1670 (7.1) |
| Other diuretics | 144 (0.6) |
| Nitrates | 251 (1.1) |
| Other hypertension drugs | 82 (0.4) |
| Digoxin | 20 (0.1) |
| Antiarrhythmics | 44 (0.2) |
| COPD or asthma meds | 2184 (9.3) |
| Statins | 4145 (17.7) |
| Other lipid lowering drugs | 1867 (8.0) |
| Antiplatelet drugs | 2081 (8.9) |
| Oral anticoagulants | 50 (0.2) |
| Heparin | 103 (0.4) |
| NSAIDs | 11 065 (47.2) |
| Oral steroid | 6885 (29.4) |
| Opioids | 6627 (28.3) |
| Antidepressants | 1014 (4.3) |
| Antipsychotics | 222 (0.9) |
| Anticonvulsants | 696 (3.0) |
| Benzodiazepines | 2871 (12.2) |
| Anxiolytics or hypnotics | 1432 (6.1) |
| Health care use | |
| Inpatient hospitalizations | |
| 0 | 21 567 (92.0) |
| 1-2 | 1801 (7.7) |
| ≥3 | 84 (0.4) |
| No. of physician visits | |
| 0-2 | 6220 (26.5) |
| 3-5 | 4169 (17.8) |
| ≥6 | 13 063 (55.7) |
| Clinical variables, mean (SD)b | |
| Body mass indexc | 26.5 (3.5) |
| Waist circumference, cm | 89.6 (8.9) |
| HbA1c, % | 6.0 (0.3) |
| Fasting blood glucose, mg/dL | 109.2 (10.5) |
| Total cholesterol, mg/dL | 205.0 (38.3) |
| HDL-C, mg/dL | 49.8 (12.7) |
| LDL-C, mg/dL | 129.9 (40.0) |
| Triglyceride, mg/dL | 186.6 (128.2) |
| Serum creatinine, mg/dL | 0.9 (0.2) |
| Estimated glomerular filtration rate, mL/min/1.73 m2 | 98.7 (14.0) |
| Systolic blood pressure, mm Hg | 124.0 (14.1) |
| Diastolic blood pressure, mm Hg | 78.8 (10.4) |
| Hemoglobin, g/dL | 15.1 (1.4) |
| AST, IU/L | 31.6 (20.3) |
| ALT, IU/L | 43.1 (35.3) |
| GGT, IU/L | 60.8 (66.6) |
| Proteinuria | |
| Negative | 13 377 (57.0) |
| Trace | 1049 (4.5) |
| ≥1 | 338 (1.4) |
| ≥2 | 137 (0.6) |
| ≥3 | 38 (0.2) |
| ≥4 | 8 (0.0) |
| Unknown | 8505 (36.3) |
Abbreviations: ACE, angiotensin converting enzyme; ALT, alanine transaminase; ARB, angiotensin receptor blocker; AST, aspartate transaminase; COPD, chronic obstructive pulmonary disease; FPG, fasting plasma glucose; GGT, gamma glutamyl transferase; HbA1c, glycated hemoglobin A1c; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; NSAIDs, nonsteroidal anti-inflammatory drugs.
To convert alanine transaminase to microkatals per liter, multiply by 0.0167; aspartate transaminase to microkatals per liter, multiply by 0.0167; creatinine to micromoles per liter, multiply by 88.4; gamma glutamyl transferase to microkatals per liter, multiply by 0.0167; glucose, multiply by 0.0555; high-density lipoprotein cholesterol to millimoles per liter, multiply by 0.259; low-density lipoprotein cholesterol to millimoles per liter, multiply by 0.259; total cholesterol to millimoles per liter, multiply by 0.259; triglycerides to millimoles per liter, multiply by 0.113.
Assessed within 1 year before the cohort entry date unless otherwise specified.
Assessed within 3 years before cohort entry date.
Calculated as weight in kilograms divided by height in meters squared.
Risk of 5-Year MACE and All-Cause Mortality
Table 2 presents the risk of MACE and all-cause mortality associated with the timing of the ADM initiation. The 5-year absolute risk of MACE progressively increased from 0.46% (95% CI, 0.21% to 1.54%) to 1.43% (95% CI, 1.28% to 1.58%) with delayed treatment initiation. Compared with no initiation within 12 months, earlier initiation showed lower point estimates for MACE with the 95% CI crossing 1 (ADM initiation within 3 months: RR, 0.32; 95% CI, 0.15-1.11; within 6 months: RR, 0.65; 95% CI, 0.41-1.29; within 12 months: RR, 0.93; 95% CI, 0.70-1.41). Corresponding RDs were −0.97% (95% CI, −1.26% to 0.14%), −0.49% (−0.84% to 0.40%), and −0.10% (−0.44% to 0.57%), respectively. For 5-year all-cause mortality, ADM initiation within 3 months showed the lowest absolute risk (0.18%; 95% CI, 0.06% to 0.55%), while initiation within 6 months (0.40%; 95% CI, 0.18% to 0.62%) and within 12 months (0.37%; 95% CI, 0.22% to 0.52%) yielded similar risks, and no initiation within 12 months showed the highest risk (0.58%; 95% CI, 0.49% to 0.69%). ADM initiation within 3 months yielded significant risk reduction for all-cause mortality compared with the control in both relative (RR, 0.31; 95% CI, 0.10-0.98) and absolute (RD, −0.40%; 95% CI, −0.57% to −0.01%) scales. Compared with no initiation within 12 months, lowered risks of mortality were observed for all strategies (within 3 months: RR, 0.31; 95% CI, 0.10-0.98; within 6 months: RR, 0.69; 95% CI, 0.29-1.09; within 12 months: RR, 0.64; 95% CI, 0.37-0.92). Survival probability curves presented consistent results with the effect estimates (Figure 2). The distribution of stabilized inverse probability of censoring weights is presented in eTable 10 in Supplement 1.
Table 2. Risk of Major Cardiovascular Events and All-Cause Mortality Associated With the Timing of the Antidiabetic Medication Initiation.
| Outcomes and treatment strategies | Events/patients, No. (%) | Total person-years | 5 y risk (95% CI) | ||
|---|---|---|---|---|---|
| Absolute risk, % | Risk difference, % | Risk ratio | |||
| Major adverse cardiovascular events | |||||
| Strategy 1 (within 3 mo) | 35/23 452 (0.15) | 15 880 | 0.46 (0.21 to 1.54) | −0.97 (−1.26 to 0.14) | 0.32 (0.15 to 1.11) |
| Strategy 2 (within 6 mo) | 68/23 452 (0.29) | 23 587 | 0.93 (0.58 to 1.80) | −0.49 (−0.84 to 0.40) | 0.65 (0.41 to 1.29) |
| Strategy 3 (within 12 mo) | 107/23 452 (0.46) | 35 904 | 1.32 (1.03 to 2.00) | −0.10 (−0.44 to 0.57) | 0.93 (0.70 to 1.41) |
| Control (no initiation) | 191/23 452 (0.52) | 68 190 | 1.43 (1.28 to 1.58) | 0 [Reference] | 1 [Reference] |
| All-cause mortality | |||||
| Strategy 1 (within 3 mo) | 14/23 452 (0.06) | 15 918 | 0.18 (0.06 to 0.55) | −0.40 (−0.57 to −0.01) | 0.31 (0.10 to 0.98) |
| Strategy 2 (within 6 mo) | 28/23 452 (0.12) | 23 670 | 0.40 (0.18 to 0.62) | −0.18 (−0.44 to 0.05) | 0.69 (0.29 to 1.09) |
| Strategy 3 (within 12 mo) | 39/23 452 (0.17) | 36 033 | 0.37 (0.22 to 0.52) | −0.21 (−0.40 to −0.04) | 0.64 (0.37 to 0.92) |
| Control (no initiation) | 77/23 452 (0.33) | 68 436 | 0.58 (0.49 to 0.69) | 0 [Reference] | 1 [Reference] |
Figure 2. Survival Probability Chart of 5-Year Major Adverse Cardiovascular Events and All-Cause Mortality.
The number at risk was estimated based on the crude (unweighted) population.
Results of Subgroup and Sensitivity Analysis
Earlier initiation of ADM showed greater cardiovascular benefit among patients who were younger than 65 years, female, had prevalent cardiovascular disease, entered the cohort between 2018 and 2022, or had either baseline glycemic status. In contrast, among participants without baseline cardiovascular disease, initiation earlier than 12 months was not associated with reduced risk compared with no initiation within 12 months (eTable 11 and eTable 12 in Supplement 1). Modeling insulin use and proxies of health care-seeking behavior as time-varying covariates and applying truncation for inverse probability of censoring weights yielded findings consistent with the main analysis (eTables 13-15 in Supplement 1). When censoring at the time of ADM discontinuation, we observed progressively lower risks of MACE and mortality with earlier initiation, with more pronounced risk reductions for MACE than in the main analysis (eTable 16 in Supplement 1). In the negative control outcome analyses, no statistically significant associations were observed between the timing of ADM initiation and the risk of herpes zoster infection or all-cause injury (eTable 17 in Supplement 1). In the landmark analysis, the treatment group showed a lower trend of risks compared with the control group, consistent with the clone-censor-weight analysis, although the effect size was attenuated (eTable 18 in Supplement 1).
Discussion
In this target trial emulation study, initiating ADM within 12 months of exceeding the T2D diagnostic threshold was associated with lower 5-year risk of all-cause mortality. The association between earlier initiation and lower risk was pronounced among individuals with prevalent cardiovascular disease. Sensitivity analyses supported the robustness of these findings and landmark analysis suggested that the clone-censor-weight approach minimized time-related and selection biases.
Current clinical guidelines recommend immediate treatment initiation at the time T2D is diagnosed.4 Nevertheless, our study demonstrated that approximately 50% of participants did not initiate ADM during the 5-year follow-up period (eTable 2 in Supplement 1). Indeed, the disease awareness rate for T2D is known to be approximately 60%,17,18 and given the asymptomatic nature of T2D, where diagnosis relies on abnormal laboratory values rather than patient symptoms, patients in the early stages of the disease are susceptible to delayed treatment initiation.19,20 However, prolonged exposure to hyperglycemia exceeding 1 year is known to be associated with increased risk of irreversible long-term adverse outcomes.10 Our findings corroborate this temporal association, demonstrating that the relative risk estimates for MACE were highest when ADM initiation occurred within 3 and 6 months after the cohort entry date, with the protective effect diminishing progressively as initiation was delayed, approaching unity at 12 months. Nonetheless, initiation timing may still vary within the same strategy window, particularly in the control group, which included both late initiators after 12 months and those untreated during follow-up. Future studies with more granular timing definitions are needed to better identify the optimal initiation window after T2D diagnosis. In parallel, health care authorities should implement evidence-based practices or strategies to improve early detection of the disease and actively promote timely treatment initiation to optimize patient outcomes in T2D management.21
Notably, our findings demonstrate that the subgroup with prevalent cardiovascular disease exhibited greater risk reduction for MACE and mortality with earlier ADM initiation. Previous studies among populations with established cardiovascular disease have yielded conflicting results, partly depending on whether participants had long-standing or early-stage T2D.22 In contrast to the UKPDS trial, which showed a sustained legacy effect among early-stage T2D, several trials evaluating intensive glycemic control in individuals with long-standing T2D and established cardiovascular disease (ie, ADVANCE,23 ACCORD,24 and VADT25) reported no significant benefit—or even an increased risk—of MACE or mortality. That is, once patients have experienced chronic exposure to hyperglycemia and reached an irreversible state, subsequent glycemic control appears ineffective in reducing the risk of cardiovascular events and mortality. Moreover, the aforementioned trials did not include novel ADM with proven cardiovascular efficacy, such as GLP-1 receptor agonists26 and SGLT2 inhibitors.27 Given that these agents have been shown to significantly reduce cardiovascular event risk in patients with T2D and preexisting cardiovascular disease through multifaceted mechanisms independent of glycemic control,28,29 their early use in the course of T2D may provide long-term benefits in preventing MACE and mortality by combining these pleiotropic effects with the legacy effect of minimized hyperglycemic exposure. As these agents were included in our ADM strategy, the observed benefit of earlier ADM initiation in the prevalent cardiovascular subgroup may be interpreted in this context. Furthermore, patients entering the cohort between 2018 and 2022 showed greater cardiovascular risk reduction with earlier ADM initiation, coinciding with the introduction of SGLT2 inhibitors and GLP-1 receptor agonists for patients with T2D and established cardiovascular disease,30 after which these agents were more widely used in Korea.31,32 Subsequent studies evaluating the optimal initiation timing of individual ADM classes would help clarify drug-specific effects and guide clinical decision-making.
Avoiding hyperglycemia through timely diagnosis and treatment initiation for T2D has been emphasized since the publication of UKPDS study results.33 There, over 80 724 person-years of follow-up, early intensive glycemic control with sulfonylurea or insulin showed risk reduction of 10% for all-cause mortality, and 17% for myocardial infarction compared with conventional glycemic control (ie, diet and lifestyle modifications).9 The pathophysiological mechanisms underlying these benefits are still under investigation, but may be attributed to preventing accumulation of advanced glycation end products, elevated oxidative stress, and epigenetic modifications that upregulate proinflammatory gene expression pathways.34,35 Subsequent observational studies have sought to evaluate this legacy effect, with some suggesting that the long-term risk of macrovascular outcomes may vary across different durations and magnitude of uncontrolled hyperglycemia10,36 and others extending the evaluation of the legacy effect to include newer ADM not assessed in previous clinical trials.37 However, these studies employed designs that did not permit causal inference between treatment assignment and outcomes and were limited in their ability to account for depletion of participants during the exposure assessment period.
Our study evaluated how the specific timing of ADM initiation after exposure to hyperglycemia is associated with the long-term risk of clinical end points. Since treatment strategies cannot be defined at time 0, we emulated the target trial using the clone-censor-weight approach to eliminate immortal time and addressed selection bias. Landmark analysis, in contrasts, limits inference to survivors of the landmark period, potentially conferring an artificial survival advantage.38 The smaller effect estimates observed in the landmark analysis support this interpretation.
Limitations
Our study has several limitations that warrant consideration. First, since inverse probability of censoring weights were estimated using logistic regression models conditional on measured prognostic factors, residual selection bias due to unmeasured prognostic factors cannot be excluded. For example, time-updated information on blood pressure and lipid control was not available, precluding assessment of postbaseline control. However, null estimates from the negative control outcome analysis supported the robustness of the main findings. Second, the risk estimates may be sensitive to the weights, due to the low incidence of events and the uneven distribution of inverse probability of censoring weights across treatment strategies; therefore, the results should be interpreted with caution. Third, our study population may not be fully representative of the general population. Because eligibility required health screening results, the cohort likely reflects individuals with greater health-seeking behavior. Furthermore, the relatively younger age of the KSHS population may have contributed to a lower baseline risk of the outcomes. Fourth, our definition of the cohort entry date relied on a single measurement of HbA1c or FPG level from annual or biennial health screening results. Therefore, information for HbA1c or FPG levels between screening intervals was not available, and the true onset of T2D may have preceded the cohort entry date among individuals who had not undergone screening for several years. Nevertheless, this feature of our study reflects clinical practice, in which patients become aware of their hyperglycemic status only upon receiving screening results. Additionally, given the limited sample size, results from subgroup analyses should be interpreted cautiously and not overstated. Further studies with larger populations and complete laboratory values would be necessary to obtain additional evidence for earlier initiation of ADM and to evaluate drug-specific effects.
Conclusions
Among individuals exceeding the diagnostic threshold for T2D, earlier initiation of ADM to resolve hyperglycemia showed progressively lower risk for mortality, supporting the legacy effect observed in prior studies. Larger population-based research is needed to confirm these findings and clarify the optimal timing of novel glucose-lowering agents, particularly in patients with established cardiovascular disease.
eAppendix 1. Description for Linkage of Kangbuk Samsung Health Study (KSHS) Cohort to National Health Insurance Service (NHIS) Database
eReferences.
eAppendix 2. Overall Description of Clone-Censor-Weight Approach
eAppendix 3. Details of Baseline and Time-Varying Covariates
eAppendix 4. Details of Sensitivity Analyses
eTable 1. Antidiabetic Medications Included in This Study
eTable 2. Pattern of Treatment Initiation Over 5-Year Follow-Up
eTable 3. Proportions of Each Antidiabetic Medication Class Among Patients who Initiated Within 12 Months
eTable 4. Proportions of Each Antidiabetic Medication Class Among Patients Who Initiated Antidiabetic Medication as Monotherapy Within 12 Months
eTable 5. Definitions for Outcomes
eTable 6. Specification and Emulation of a Target Trial
eTable 7. Definitions for Comorbidities and Comedications
eTable 8. Baseline Characteristics of Patients Adhering to Assigned Treatment Strategy at Cohort Entry Date
eTable 9. Time to Initiation and Duration of Antidiabetic Medication Use Across Treatment Strategies
eTable 10. Distribution of Inverse Probability of Censoring Weight After Stabilization
eTable 11. Results of Subgroup Analysis for the Risk of 5-Year Major Adverse Cardiovascular Events
eTable 12. Results of Subgroup Analysis for the Risk of 5-Year All-Cause Mortality
eTable 13. Results of Sensitivity Analysis Assessing Insulin Use as a Time-Varying Covariate
eTable 14. Results of Sensitivity Analysis Assessing Proxies for Health Seeking Behaviors as Time-Varying Covariates
eTable 15. Results of Sensitivity Analysis Applying Truncated Inverse Probability of Censoring Weight
eTable 16. Results of Sensitivity Analysis With Additional Censoring at Time of Antidiabetic Medication Discontinuation
eTable 17. Results of Sensitivity Analysis Assessing Herpes Zoster Virus Infection and All Cause-Injury as Negative Control Outcomes
eTable 18. Results of Landmark Analysis
eFigure. Distribution of the Duration Between the Cohort Entry Date (Time-Zero) and Antidiabetic Medication Initiation Within 12 Months
Data Sharing Statement
References
- 1.Sarwar N, Gao P, Seshasai SR, et al. ; Emerging Risk Factors Collaboration . Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet. 2010;375(9733):2215-2222. doi: 10.1016/S0140-6736(10)60484-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zheng Y, Ley SH, Hu FB. Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nat Rev Endocrinol. 2018;14(2):88-98. doi: 10.1038/nrendo.2017.151 [DOI] [PubMed] [Google Scholar]
- 3.Rao Kondapally Seshasai S, Kaptoge S, Thompson A, et al. ; Emerging Risk Factors Collaboration . Diabetes mellitus, fasting glucose, and risk of cause-specific death. N Engl J Med. 2011;364(9):829-841. doi: 10.1056/NEJMoa1008862 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.American Diabetes Association Professional Practice Committee . 2. Diagnosis and classification of diabetes: standards of care in diabetes-2025. Diabetes Care. 2025;48(1)(suppl 1):S27-S49. doi: 10.2337/dc25-S002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.The Lancet Diabetes Endocrinology . Undiagnosed type 2 diabetes: an invisible risk factor. Lancet Diabetes Endocrinol. 2024;12(4):215. doi: 10.1016/S2213-8587(24)00072-X [DOI] [PubMed] [Google Scholar]
- 6.Khunti K, Wolden ML, Thorsted BL, Andersen M, Davies MJ. Clinical inertia in people with type 2 diabetes: a retrospective cohort study of more than 80,000 people. Diabetes Care. 2013;36(11):3411-3417. doi: 10.2337/dc13-0331 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589. doi: 10.1056/NEJMoa0806470 [DOI] [PubMed] [Google Scholar]
- 8.Lachin JM, Nathan DM; DCCT/EDIC Research Group . Understanding metabolic memory: the prolonged influence of glycemia during the diabetes control and complications trial (DCCT) on future risks of complications during the study of the epidemiology of diabetes interventions and complications (EDIC). Diabetes Care. 2021;44(10):2216-2224. doi: 10.2337/dc20-3097 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Adler AI, Coleman RL, Leal J, Whiteley WN, Clarke P, Holman RR. Post-trial monitoring of a randomised controlled trial of intensive glycaemic control in type 2 diabetes extended from 10 years to 24 years (UKPDS 91). Lancet. 2024;404(10448):145-155. doi: 10.1016/S0140-6736(24)00537-3 [DOI] [PubMed] [Google Scholar]
- 10.Laiteerapong N, Ham SA, Gao Y, et al. The legacy effect in type 2 diabetes: impact of early glycemic control on future complications (the diabetes & aging study). Diabetes Care. 2019;42(3):416-426. doi: 10.2337/dc17-1144 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Writing Team for the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Research Group . Sustained effect of intensive treatment of type 1 diabetes mellitus on development and progression of diabetic nephropathy: the Epidemiology of Diabetes Interventions and Complications (EDIC) study. JAMA. 2003;290(16):2159-2167. doi: 10.1001/jama.290.16.2159 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hernán MA, Sterne JAC, Higgins JPT, Shrier I, Hernández-Díaz S. A structural description of biases that generate immortal time. Epidemiology. 2025;36(1):107-114. doi: 10.1097/EDE.0000000000001808 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Gaber CE, Ghazarian AA, Strassle PD, et al. De-mystifying the clone-censor-weight method for causal research using observational data: a primer for cancer researchers. Cancer Med. 2024;13(23):e70461. doi: 10.1002/cam4.70461 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cashin AG, Hansford HJ, Hernán MA, et al. Transparent Reporting of Observational Studies Emulating a Target Trial-The TARGET Statement. JAMA. 2025;334(12):1084-1093. doi: 10.1001/jama.2025.13350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Groenwold RH, White IR, Donders AR, Carpenter JR, Altman DG, Moons KG. Missing covariate data in clinical research: when and when not to use the missing-indicator method for analysis. CMAJ. 2012;184(11):1265-1269. doi: 10.1503/cmaj.110977 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481. doi: 10.1080/01621459.1958.10501452 [DOI] [Google Scholar]
- 17.Shahrestanaki E, Mohammadian Khonsari N, Seif E, et al. The worldwide trend in diabetes awareness, treatment, and control from 1985 to 2022: a systematic review and meta-analysis of 233 population-representative studies. Front Public Health. 2024;12:1305304. doi: 10.3389/fpubh.2024.1305304 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hu X, Fang X, Wu M. Prevalence, awareness, treatment and control of type 2 diabetes in southeast China: a population-based study. J Diabetes Investig. 2024;15(8):1034-1041. doi: 10.1111/jdi.14213 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Phillips LS, Branch WT, Cook CB, et al. Clinical inertia. Ann Intern Med. 2001;135(9):825-834. doi: 10.7326/0003-4819-135-9-200111060-00012 [DOI] [PubMed] [Google Scholar]
- 20.Gopalan A, Mishra P, Alexeeff SE, et al. Prevalence and predictors of delayed clinical diagnosis of type 2 diabetes: a longitudinal cohort study. Diabet Med. 2018;35(12):1655-1662. doi: 10.1111/dme.13808 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Powell RE, Zaccardi F, Beebe C, et al. Strategies for overcoming therapeutic inertia in type 2 diabetes: a systematic review and meta-analysis. Diabetes Obes Metab. 2021;23(9):2137-2154. doi: 10.1111/dom.14455 [DOI] [PubMed] [Google Scholar]
- 22.Boussageon R, Bejan-Angoulvant T, Saadatian-Elahi M, et al. Effect of intensive glucose lowering treatment on all cause mortality, cardiovascular death, and microvascular events in type 2 diabetes: meta-analysis of randomised controlled trials. BMJ. 2011;343:d4169. doi: 10.1136/bmj.d4169 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Patel A, MacMahon S, Chalmers J, et al. ; ADVANCE Collaborative Group . Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med. 2008;358(24):2560-2572. doi: 10.1056/NEJMoa0802987 [DOI] [PubMed] [Google Scholar]
- 24.Gerstein HC, Miller ME, Byington RP, et al. ; Action to Control Cardiovascular Risk in Diabetes Study Group . Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med. 2008;358(24):2545-2559. doi: 10.1056/NEJMoa0802743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Duckworth W, Abraira C, Moritz T, et al. ; VADT Investigators . Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med. 2009;360(2):129-139. doi: 10.1056/NEJMoa0808431 [DOI] [PubMed] [Google Scholar]
- 26.Marx N, Husain M, Lehrke M, Verma S, Sattar N. GLP-1 receptor agonists for the reduction of atherosclerotic cardiovascular risk in patients with type 2 diabetes. Circulation. 2022;146(24):1882-1894. doi: 10.1161/CIRCULATIONAHA.122.059595 [DOI] [PubMed] [Google Scholar]
- 27.Patel SM, Kang YM, Im K, et al. Sodium-glucose cotransporter-2 inhibitors and major adverse cardiovascular outcomes: a SMART-C collaborative meta-analysis. Circulation. 2024;149(23):1789-1801. doi: 10.1161/CIRCULATIONAHA.124.069568 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Cowie MR, Fisher M. SGLT2 inhibitors: mechanisms of cardiovascular benefit beyond glycaemic control. Nat Rev Cardiol. 2020;17(12):761-772. doi: 10.1038/s41569-020-0406-8 [DOI] [PubMed] [Google Scholar]
- 29.Ussher JR, Drucker DJ. Glucagon-like peptide 1 receptor agonists: cardiovascular benefits and mechanisms of action. Nat Rev Cardiol. 2023;20(7):463-474. doi: 10.1038/s41569-023-00849-3 [DOI] [PubMed] [Google Scholar]
- 30.Davies MJ, D’Alessio DA, Fradkin J, et al. Management of hyperglycemia in type 2 diabetes, 2018. A consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetes Care. 2018;41(12):2669-2701. doi: 10.2337/dci18-0033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Ha KH, Shin S, Na E, Kim DJ. Trends in prescribing sodium-glucose cotransporter 2 inhibitors for individuals with type 2 diabetes with and without cardiovascular-renal disease in South Korea, 2015-2021. J Diabetes Investig. 2025;16(2):215-224. doi: 10.1111/jdi.14363 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Eom YR, Joo H, Chae SE, Je NK. Prescribing patterns of SGLT2 inhibitors and GLP-1 receptor agonists in patients with T2DM and ASCVD in South Korea. Pharmacoepidemiol Drug Saf. 2025;34(7):e70183. doi: 10.1002/pds.70183 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.UK Prospective Diabetes Study (UKPDS) Group . Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352(9131):837-853. doi: 10.1016/S0140-6736(98)07019-6 [DOI] [PubMed] [Google Scholar]
- 34.Reddy MA, Zhang E, Natarajan R. Epigenetic mechanisms in diabetic complications and metabolic memory. Diabetologia. 2015;58(3):443-455. doi: 10.1007/s00125-014-3462-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Haythorne E, Lloyd M, Walsby-Tickle J, et al. Altered glycolysis triggers impaired mitochondrial metabolism and mTORC1 activation in diabetic β-cells. Nat Commun. 2022;13(1):6754. doi: 10.1038/s41467-022-34095-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Ceriello A, Lucisano G, Prattichizzo F, et al. ; AMD Annals study group . The legacy effect of hyperglycemia and early use of SGLT-2 inhibitors: a cohort study with newly-diagnosed people with type 2 diabetes. Lancet Reg Health Eur. 2023;31:100666. doi: 10.1016/j.lanepe.2023.100666 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Deng S, Zhao H, Chai S, et al. Influence of early use of sodium-glucose transport protein 2 inhibitors, glucagon-like peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors on the legacy effect of hyperglycemia. Front Endocrinol (Lausanne). 2024;15:1369908. doi: 10.3389/fendo.2024.1369908 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Duchesneau ED, Jackson BE, Webster-Clark M, et al. The timing, the treatment, the question: comparison of epidemiologic approaches to minimize immortal time bias in real-world data using a surgical oncology example. Cancer Epidemiol Biomarkers Prev. 2022;31(11):2079-2086. doi: 10.1158/1055-9965.EPI-22-0495 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eAppendix 1. Description for Linkage of Kangbuk Samsung Health Study (KSHS) Cohort to National Health Insurance Service (NHIS) Database
eReferences.
eAppendix 2. Overall Description of Clone-Censor-Weight Approach
eAppendix 3. Details of Baseline and Time-Varying Covariates
eAppendix 4. Details of Sensitivity Analyses
eTable 1. Antidiabetic Medications Included in This Study
eTable 2. Pattern of Treatment Initiation Over 5-Year Follow-Up
eTable 3. Proportions of Each Antidiabetic Medication Class Among Patients who Initiated Within 12 Months
eTable 4. Proportions of Each Antidiabetic Medication Class Among Patients Who Initiated Antidiabetic Medication as Monotherapy Within 12 Months
eTable 5. Definitions for Outcomes
eTable 6. Specification and Emulation of a Target Trial
eTable 7. Definitions for Comorbidities and Comedications
eTable 8. Baseline Characteristics of Patients Adhering to Assigned Treatment Strategy at Cohort Entry Date
eTable 9. Time to Initiation and Duration of Antidiabetic Medication Use Across Treatment Strategies
eTable 10. Distribution of Inverse Probability of Censoring Weight After Stabilization
eTable 11. Results of Subgroup Analysis for the Risk of 5-Year Major Adverse Cardiovascular Events
eTable 12. Results of Subgroup Analysis for the Risk of 5-Year All-Cause Mortality
eTable 13. Results of Sensitivity Analysis Assessing Insulin Use as a Time-Varying Covariate
eTable 14. Results of Sensitivity Analysis Assessing Proxies for Health Seeking Behaviors as Time-Varying Covariates
eTable 15. Results of Sensitivity Analysis Applying Truncated Inverse Probability of Censoring Weight
eTable 16. Results of Sensitivity Analysis With Additional Censoring at Time of Antidiabetic Medication Discontinuation
eTable 17. Results of Sensitivity Analysis Assessing Herpes Zoster Virus Infection and All Cause-Injury as Negative Control Outcomes
eTable 18. Results of Landmark Analysis
eFigure. Distribution of the Duration Between the Cohort Entry Date (Time-Zero) and Antidiabetic Medication Initiation Within 12 Months
Data Sharing Statement
